DonorsChoose¶
DonorsChoose.org receives hundreds of thousands of project proposals each year for classroom projects in need of funding. Right now, a large number of volunteers is needed to manually screen each submission before it's approved to be posted on the DonorsChoose.org website.
Next year, DonorsChoose.org expects to receive close to 500,000 project proposals. As a result, there are three main problems they need to solve:
- How to scale current manual processes and resources to screen 500,000 projects so that they can be posted as quickly and as efficiently as possible
- How to increase the consistency of project vetting across different volunteers to improve the experience for teachers
- How to focus volunteer time on the applications that need the most assistance
The goal of the competition is to predict whether or not a DonorsChoose.org project proposal submitted by a teacher will be approved, using the text of project descriptions as well as additional metadata about the project, teacher, and school. DonorsChoose.org can then use this information to identify projects most likely to need further review before approval.
About the DonorsChoose Data Set¶
The train.csv data set provided by DonorsChoose contains the following features:
| Feature | Description |
|---|---|
project_id |
A unique identifier for the proposed project. Example: p036502 |
project_title |
Title of the project. Examples:
|
project_grade_category |
Grade level of students for which the project is targeted. One of the following enumerated values:
|
project_subject_categories |
One or more (comma-separated) subject categories for the project from the following enumerated list of values:
Examples:
|
school_state |
State where school is located (Two-letter U.S. postal code). Example: WY |
project_subject_subcategories |
One or more (comma-separated) subject subcategories for the project. Examples:
|
project_resource_summary |
An explanation of the resources needed for the project. Example:
|
project_essay_1 |
First application essay* |
project_essay_2 |
Second application essay* |
project_essay_3 |
Third application essay* |
project_essay_4 |
Fourth application essay* |
project_submitted_datetime |
Datetime when project application was submitted. Example: 2016-04-28 12:43:56.245 |
teacher_id |
A unique identifier for the teacher of the proposed project. Example: bdf8baa8fedef6bfeec7ae4ff1c15c56 |
teacher_prefix |
Teacher's title. One of the following enumerated values:
|
teacher_number_of_previously_posted_projects |
Number of project applications previously submitted by the same teacher. Example: 2 |
* See the section Notes on the Essay Data for more details about these features.
Additionally, the resources.csv data set provides more data about the resources required for each project. Each line in this file represents a resource required by a project:
| Feature | Description |
|---|---|
id |
A project_id value from the train.csv file. Example: p036502 |
description |
Desciption of the resource. Example: Tenor Saxophone Reeds, Box of 25 |
quantity |
Quantity of the resource required. Example: 3 |
price |
Price of the resource required. Example: 9.95 |
Note: Many projects require multiple resources. The id value corresponds to a project_id in train.csv, so you use it as a key to retrieve all resources needed for a project:
The data set contains the following label (the value you will attempt to predict):
| Label | Description |
|---|---|
project_is_approved |
A binary flag indicating whether DonorsChoose approved the project. A value of 0 indicates the project was not approved, and a value of 1 indicates the project was approved. |
Notes on the Essay Data¶
-
Prior to May 17, 2016, the prompts for the essays were as follows:
- __project_essay_1:__ "Introduce us to your classroom"
- __project_essay_2:__ "Tell us more about your students"
- __project_essay_3:__ "Describe how your students will use the materials you're requesting"
- __project_essay_3:__ "Close by sharing why your project will make a difference"
-
Starting on May 17, 2016, the number of essays was reduced from 4 to 2, and the prompts for the first 2 essays were changed to the following:
- __project_essay_1:__ "Describe your students: What makes your students special? Specific details about their background, your neighborhood, and your school are all helpful."
- __project_essay_2:__ "About your project: How will these materials make a difference in your students' learning and improve their school lives?"
For all projects with project_submitted_datetime of 2016-05-17 and later, the values of project_essay_3 and project_essay_4 will be NaN.
In [1]:
%matplotlib inline
import warnings
warnings.filterwarnings("ignore")
import sqlite3
import pandas as pd
import numpy as np
import nltk
import string
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.feature_extraction.text import TfidfTransformer
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.metrics import confusion_matrix
from sklearn import metrics
from sklearn.metrics import roc_curve, auc
from nltk.stem.porter import PorterStemmer
import re
# Tutorial about Python regular expressions: https://pymotw.com/2/re/
import string
from nltk.corpus import stopwords
from nltk.stem import PorterStemmer
from nltk.stem.wordnet import WordNetLemmatizer
from gensim.models import Word2Vec
from gensim.models import KeyedVectors
import pickle
from tqdm import tqdm
import os
from plotly import plotly
import plotly.offline as offline
import plotly.graph_objs as go
offline.init_notebook_mode()
from collections import Counter
Assignment 9: RF and GBDT¶
Response Coding: Example¶
The response tabel is built only on train dataset. For a category which is not there in train data and present in test data, we will encode them with default values Ex: in our test data if have State: D then we encode it as [0.5, 0.05]
- Apply both Random Forrest and GBDT on these feature sets
- Set 1: categorical(instead of one hot encoding, try response coding: use probability values), numerical features + project_title(BOW) + preprocessed_eassay (BOW)
- Set 2: categorical(instead of one hot encoding, try response coding: use probability values), numerical features + project_title(TFIDF)+ preprocessed_eassay (TFIDF)
- Set 3: categorical(instead of one hot encoding, try response coding: use probability values), numerical features + project_title(AVG W2V)+ preprocessed_eassay (AVG W2V)
- Set 4: categorical(instead of one hot encoding, try response coding: use probability values), numerical features + project_title(TFIDF W2V)+ preprocessed_eassay (TFIDF W2V)
- The hyper paramter tuning (Consider any two hyper parameters preferably n_estimators, max_depth)
- Consider the following range for hyperparameters n_estimators = [10, 50, 100, 150, 200, 300, 500, 1000], max_depth = [2, 3, 4, 5, 6, 7, 8, 9, 10]
- Find the best hyper parameter which will give the maximum AUC value
- find the best hyper paramter using k-fold cross validation/simple cross validation data
- use gridsearch cv or randomsearch cv or you can write your own for loops to do this task
-
Representation of results
- You need to plot the performance of model both on train data and cross validation data for each hyper parameter, like shown in the figure
with X-axis as n_estimators, Y-axis as max_depth, and Z-axis as AUC Score , we have given the notebook which explains how to plot this 3d plot, you can find it in the same drive 3d_scatter_plot.ipynb - You need to plot the performance of model both on train data and cross validation data for each hyper parameter, like shown in the figure
seaborn heat maps with rows as n_estimators, columns as max_depth, and values inside the cell representing AUC Score - You can choose either of the plotting techniques: 3d plot or heat map
- Once after you found the best hyper parameter, you need to train your model with it, and find the AUC on test data and plot the ROC curve on both train and test.
- Along with plotting ROC curve, you need to print the confusion matrix with predicted and original labels of test data points
or
- You need to plot the performance of model both on train data and cross validation data for each hyper parameter, like shown in the figure
- Conclusion
- You need to summarize the results at the end of the notebook, summarize it in the table format. To print out a table please refer to this prettytable library link
- You need to summarize the results at the end of the notebook, summarize it in the table format. To print out a table please refer to this prettytable library link
Note: Data Leakage
- There will be an issue of data-leakage if you vectorize the entire data and then split it into train/cv/test.
- To avoid the issue of data-leakage, make sure to split your data first and then vectorize it.
- While vectorizing your data, apply the method fit_transform() on you train data, and apply the method transform() on cv/test data.
- For more details please go through this link.
1.1 Reading Data¶
In [2]:
project_data=pd.read_csv('train_data.csv', nrows=35000)
resource_data=pd.read_csv('resources.csv')
In [3]:
print("number of data points in train data", project_data.shape)
print('-'*50)
print("the attributes of data :", project_data.columns.values)
In [4]:
print("Number of data points in train data", resource_data.shape)
print(resource_data.columns.values)
resource_data.head(2)
Out[4]:
In [5]:
# https://stackoverflow.com/questions/22407798/how-to-reset-a-dataframes-indexes-for-all-groups-in-one-step
price_data=resource_data.groupby('id').agg({'price':'sum','quantity':'sum'}).reset_index()
price_data.head(2)
Out[5]:
In [6]:
# join two dataframes in python:
project_data=pd.merge(project_data, price_data, on='id', how='left')
In [7]:
project_data.head(2)
Out[7]:
In [8]:
# presence of the numerical digits in a strings with numeric : https://stackoverflow.com/a/19859308/8089731
def hasNumbers(inputString):
return any(i.isdigit()for i in inputString)
p1=project_data[['id','project_resource_summary']]
p1=pd.DataFrame(data=p1)
p1.columns=['id','digits_in_summary']
p1['digits_in_summary']=p1['digits_in_summary'].map(hasNumbers)
# https://stackoverflow.com/a/17383325/8089731
p1['digits_in_summary'] = p1['digits_in_summary'].astype(int)
project_data=pd.merge(project_data,p1,on='id',how='left')
project_data.head(5)
Out[8]:
1.2 preprocessing of project_subject_categories¶
In [9]:
categories=list(project_data['project_subject_categories'].values)
# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039
# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/
# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string
# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python
cat_list=[]
for i in categories:
temp=""
# consider we have text like this "Math & Science, Warmth, Care & Hunger"
for j in i.split(','):# it will split it in three parts ["Math & Science", "Warmth", "Care & Hunger"]
if 'The' in j.split():# this will split each of the catogory based on space "Math & Science"=> "Math","&", "Science"
j=j.replace('The','')# if we have the words "The" we are going to replace it with ''(i.e removing 'The')
j=j.replace(' ','')# we are placeing all the ' '(space) with ''(empty) ex:"Math & Science"=>"Math&Science"
temp+=j.strip()+" " #" abc ".strip() will return "abc", remove the trailing spaces
temp=temp.replace('&','_')# we are replacing the & value into
cat_list.append(temp.strip())
project_data['clean_categories']=cat_list
project_data.drop(['project_subject_categories'], axis=1, inplace=True)
project_data.head(5)
Out[9]:
In [10]:
# count of all the words in corpus python: https://stackoverflow.com/a/22898595/4084039
from collections import Counter
my_counter = Counter()
for word in project_data['clean_categories'].values:
my_counter.update(word.split())
my_counter
Out[10]:
In [11]:
# dict sort by value python: https://stackoverflow.com/a/613218/4084039
cat_dict = dict(my_counter)
sorted_cat_dict = dict(sorted(cat_dict.items(), key=lambda kv: kv[1]))
1.3 preprocessing of project_subject_subcategories¶
In [12]:
sub_catogories = list(project_data['project_subject_subcategories'].values)
# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039
# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/
# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string
# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python
sub_cat_list = []
for i in sub_catogories:
temp = ""
# consider we have text like this "Math & Science, Warmth, Care & Hunger"
for j in i.split(','): # it will split it in three parts ["Math & Science", "Warmth", "Care & Hunger"]
if 'The' in j.split(): # this will split each of the catogory based on space "Math & Science"=> "Math","&", "Science"
j=j.replace('The','') # if we have the words "The" we are going to replace it with ''(i.e removing 'The')
j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:"Math & Science"=>"Math&Science"
temp +=j.strip()+" "#" abc ".strip() will return "abc", remove the trailing spaces
temp = temp.replace('&','_')
sub_cat_list.append(temp.strip())
In [13]:
project_data['clean_subcategories'] = sub_cat_list
project_data.drop(['project_subject_subcategories'], axis=1, inplace=True)
project_data.head(2)
Out[13]:
In [14]:
# count of all the words in corpus python: https://stackoverflow.com/a/22898595/4084039
from collections import Counter
my_counter = Counter()
for word in project_data['clean_subcategories'].values:
my_counter.update(word.split())
In [15]:
# dict sort by value python: https://stackoverflow.com/a/613218/4084039
sub_cat_dict = dict(my_counter)
sorted_sub_cat_dict = dict(sorted(sub_cat_dict.items(), key=lambda kv: kv[1]))
1.3 Text preprocessing¶
In [16]:
# merge two column text dataframe:
project_data["essay"] = project_data["project_essay_1"].map(str) +\
project_data["project_essay_2"].map(str) + \
project_data["project_essay_3"].map(str) + \
project_data["project_essay_4"].map(str)
In [17]:
# https://stackoverflow.com/a/47091490/4084039
import re
def decontracted(phrase):
# specific
phrase = re.sub(r"won't", "will not", phrase)
phrase = re.sub(r"can\'t", "can not", phrase)
# general
phrase = re.sub(r"n\'t", " not", phrase)
phrase = re.sub(r"\'re", " are", phrase)
phrase = re.sub(r"\'s", " is", phrase)
phrase = re.sub(r"\'d", " would", phrase)
phrase = re.sub(r"\'ll", " will", phrase)
phrase = re.sub(r"\'t", " not", phrase)
phrase = re.sub(r"\'ve", " have", phrase)
phrase = re.sub(r"\'m", " am", phrase)
return phrase
In [18]:
# https://gist.github.com/sebleier/554280
# we are removing the words from the stop words list: 'no', 'nor', 'not'
stopwords= ['i', 'me', 'my', 'myself', 'we', 'our', 'ours', 'ourselves', 'you', "you're", "you've",\
"you'll", "you'd", 'your', 'yours', 'yourself', 'yourselves', 'he', 'him', 'his', 'himself', \
'she', "she's", 'her', 'hers', 'herself', 'it', "it's", 'its', 'itself', 'they', 'them', 'their',\
'theirs', 'themselves', 'what', 'which', 'who', 'whom', 'this', 'that', "that'll", 'these', 'those', \
'am', 'is', 'are', 'was', 'were', 'be', 'been', 'being', 'have', 'has', 'had', 'having', 'do', 'does', \
'did', 'doing', 'a', 'an', 'the', 'and', 'but', 'if', 'or', 'because', 'as', 'until', 'while', 'of', \
'at', 'by', 'for', 'with', 'about', 'against', 'between', 'into', 'through', 'during', 'before', 'after',\
'above', 'below', 'to', 'from', 'up', 'down', 'in', 'out', 'on', 'off', 'over', 'under', 'again', 'further',\
'then', 'once', 'here', 'there', 'when', 'where', 'why', 'how', 'all', 'any', 'both', 'each', 'few', 'more',\
'most', 'other', 'some', 'such', 'only', 'own', 'same', 'so', 'than', 'too', 'very', \
's', 't', 'can', 'will', 'just', 'don', "don't", 'should', "should've", 'now', 'd', 'll', 'm', 'o', 're', \
've', 'y', 'ain', 'aren', "aren't", 'couldn', "couldn't", 'didn', "didn't", 'doesn', "doesn't", 'hadn',\
"hadn't", 'hasn', "hasn't", 'haven', "haven't", 'isn', "isn't", 'ma', 'mightn', "mightn't", 'mustn',\
"mustn't", 'needn', "needn't", 'shan', "shan't", 'shouldn', "shouldn't", 'wasn', "wasn't", 'weren', "weren't", \
'won', "won't", 'wouldn', "wouldn't"]
In [19]:
# Combining all the above statemennts
from tqdm import tqdm
preprocessed_essays = []
# tqdm is for printing the status bar
for sentance in tqdm(project_data['essay'].values):
sent = decontracted(sentance)
sent = sent.replace('\\r', ' ')
sent = sent.replace('\\"', ' ')
sent = sent.replace('\\n', ' ')
sent = re.sub('[^A-Za-z0-9]+', ' ', sent)
# https://gist.github.com/sebleier/554280
sent = ' '.join(e for e in sent.split() if e not in stopwords)
preprocessed_essays.append(sent.lower().strip())
In [20]:
preprocessed_essays[2000]
Out[20]:
In [21]:
from tqdm import tqdm
preprocessed_titles = []
# tqdm is for printing the status bar
for title in tqdm(project_data['project_title'].values):
_title = decontracted(title)
_title = _title.replace('\\r', ' ')
_title = _title.replace('\\"', ' ')
_title = _title.replace('\\n', ' ')
_title = re.sub('[^A-Za-z0-9]+', ' ', _title)
# https://gist.github.com/sebleier/554280
_title = ' '.join(e for e in _title.split() if e not in stopwords)
preprocessed_titles.append(_title.lower().strip())
In [22]:
preprocessed_titles[2000]
Out[22]:
In [23]:
project_grade_catogories = list(project_data['project_grade_category'].values)
# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039
# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/
# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string
# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python
project_grade_cat_list = []
for i in tqdm(project_grade_catogories):
temp = ""
# consider we have text like this "Math & Science, Warmth, Care & Hunger"
for j in i.split(','): # it will split it in three parts ["Math & Science", "Warmth", "Care & Hunger"]
if 'The' in j.split(): # this will split each of the catogory based on space "Math & Science"=> "Math","&", "Science"
j=j.replace('The','') # if we have the words "The" we are going to replace it with ''(i.e removing 'The')
j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:"Math & Science"=>"Math&Science"
temp +=j.strip()+" "#" abc ".strip() will return "abc", remove the trailing spaces
temp = temp.replace('&','_')
project_grade_cat_list.append(temp.strip())
In [24]:
project_grade_cat_list[2000]
Out[24]:
In [25]:
project_data['clean_project_grade_category'] = project_grade_cat_list
project_data.drop(['project_grade_category'], axis=1, inplace=True)
project_data.head()
Out[25]:
In [26]:
project_data.drop(['project_essay_1','project_essay_2','project_essay_3','project_essay_4'], axis=1, inplace=True)
project_data.head()
Out[26]:
In [27]:
project_data['preprocessed_essays'] = preprocessed_essays
project_data['preprocessed_titles'] = preprocessed_titles
In [28]:
#Replacing Nan's with maximum occured value: https://stackoverflow.com/a/51053916/8089731
project_data['teacher_prefix'].value_counts().argmax()
project_data.fillna(value=project_data['teacher_prefix'].value_counts().argmax(),axis=1,inplace=True)
In [29]:
# School State
project_school_catogories = list(project_data['school_state'].values)
# remove special characters from list of strings python: https://stackoverflow.com/a/47301924/4084039
# https://www.geeksforgeeks.org/removing-stop-words-nltk-python/
# https://stackoverflow.com/questions/23669024/how-to-strip-a-specific-word-from-a-string
# https://stackoverflow.com/questions/8270092/remove-all-whitespace-in-a-string-in-python
project_school_cat_list = []
for i in tqdm(project_school_catogories):
temp = ""
# consider we have text like this "Math & Science, Warmth, Care & Hunger"
for j in i.split(','): # it will split it in three parts ["Math & Science", "Warmth", "Care & Hunger"]
if 'The' in j.split(): # this will split each of the catogory based on space "Math & Science"=> "Math","&", "Science"
j=j.replace('The','') # if we have the words "The" we are going to replace it with ''(i.e removing 'The')
j = j.replace(' ','') # we are placeing all the ' '(space) with ''(empty) ex:"Math & Science"=>"Math&Science"
temp +=j.strip()+" "#" abc ".strip() will return "abc", remove the trailing spaces
temp = temp.replace('&','_')
project_school_cat_list.append(temp.strip())
In [30]:
project_data['clean_project_school_category'] = project_school_cat_list
project_data.drop(['school_state'], axis=1, inplace=True)
project_data.head(2)
Out[30]:
In [31]:
from tqdm import tqdm
preprocessed_teacher_prefix = []
# tqdm is for printing the status bar
for prefix in tqdm(project_data['teacher_prefix'].values):
_prefix = decontracted(prefix)
_prefix = _prefix.replace('\\r', ' ')
_prefix = _prefix.replace('\\"', ' ')
_prefix = _prefix.replace('\\n', ' ')
_prefix = _prefix.replace('\\.', ' ')
_prefix = re.sub('[^A-Za-z0-9]+', ' ', _prefix)
# https://gist.github.com/sebleier/554280
_prefix = ' '.join(e for e in _prefix.split() if e not in stopwords)
preprocessed_teacher_prefix.append(_prefix.lower().strip())
In [32]:
preprocessed_teacher_prefix[20000]
Out[32]:
In [33]:
project_data['clean_teacher_prefix_category'] = preprocessed_teacher_prefix
project_data.drop(['teacher_prefix'], axis=1, inplace=True)
project_data.head(2)
Out[33]:
In [34]:
project_data['clean_teacher_prefix_category'][20000]
Out[34]:
1.5 Preparing data for models¶
2. Random Forest and GBDT
2.1 Splitting data into Train and cross validation(or test): Stratified Sampling
In [35]:
# please write all the code with proper documentation, and proper titles for each subsection
# go through documentations and blogs before you start coding
# first figure out what to do, and then think about how to do.
# reading and understanding error messages will be very much helpfull in debugging your code
# when you plot any graph make sure you use
# a. Title, that describes your plot, this will be very helpful to the reader
# b. Legends if needed
# c. X-axis label
# d. Y-axis label
In [36]:
project_data.columns
Out[36]:
In [37]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from sklearn.model_selection import cross_val_score
from collections import Counter
from sklearn.metrics import accuracy_score
from sklearn import model_selection
In [38]:
X_train, X_test, y_train, y_test = train_test_split(project_data,project_data['project_is_approved'], test_size=0.33, stratify = project_data['project_is_approved'])
#X_train, X_cv, y_train, y_cv = train_test_split(X_train, y_train, test_size=0.33, stratify=y_train)
print(X_train.shape, y_train.shape)
#print(X_cv.shape, y_cv.shape)
print(X_test.shape, y_test.shape)
2.2 Make Data Model Ready: encoding numerical, categorical features¶
In [39]:
# please write all the code with proper documentation, and proper titles for each subsection
# go through documentations and blogs before you start coding
# first figure out what to do, and then think about how to do.
# reading and understanding error messages will be very much helpfull in debugging your code
# make sure you featurize train and test data separatly
# when you plot any graph make sure you use
# a. Title, that describes your plot, this will be very helpful to the reader
# b. Legends if needed
# c. X-axis label
# d. Y-axis label
1. project_subject_categories: clean_categories¶
In [40]:
Xtrain_pos = X_train.loc[X_train['project_is_approved'] == 1]
Xtrain_neg = X_train.loc[X_train['project_is_approved'] == 0]
In [41]:
clean_pos_cat = {}
for a in Xtrain_pos['clean_categories'] :
for b in a.split():
if b not in clean_pos_cat :
clean_pos_cat[b] = 1
else :
clean_pos_cat[b] += 1
clean_neg_cat = {}
for a in Xtrain_neg['clean_categories'] :
for b in a.split():
if b not in clean_neg_cat :
clean_neg_cat[b] = 1
else :
clean_neg_cat[b] += 1
clean_cat_xtrain = {}
for a in X_train['clean_categories'] :
for b in a.split():
if b not in clean_cat_xtrain :
clean_cat_xtrain[b] = 1
else :
clean_cat_xtrain[b] += 1
In [42]:
pos_cat_p = {}
for p in clean_cat_xtrain.keys():
pos_cat_p[p] = (clean_pos_cat[p])/float(clean_cat_xtrain[p])
neg_cat_n = {}
for n in clean_cat_xtrain.keys():
neg_cat_n[n] = (clean_neg_cat[n])/float(clean_cat_xtrain[n])
cat_0n_xtrain = []
cat_1p_xtrain = []
for a in X_train["clean_categories"] :
b = a.split()
if len(b) == 1 :
cat_0n_xtrain.append(neg_cat_n[a])
cat_1p_xtrain.append(pos_cat_p[a])
else :
c = neg_cat_n[b[0]]
d =neg_cat_n[b[1]]
e = pos_cat_p[b[0]]
f = pos_cat_p[b[1]]
cat_0n_xtrain.append(c*d)
cat_1p_xtrain.append(e*f)
X_train["cat_0n"] = cat_0n_xtrain
X_train["cat_1p"] = cat_1p_xtrain
In [43]:
import math
cat_0n_xtest = []
cat_1p_xtest = []
for a in X_test["clean_categories"] :
b = a.split()
if len(b) == 1 :
cat_0n_xtest.append(neg_cat_n[a])
cat_1p_xtest.append(pos_cat_p[a])
else :
c = neg_cat_n[b[0]]
d = neg_cat_n[b[1]]
e = pos_cat_p[b[0]]
f = pos_cat_p[b[1]]
cat_0n_xtest.append(c*d)
cat_1p_xtest.append(e*f)
flag = 0
for i in range(len(cat_0n_xtest)):
if(math.isnan(cat_0n_xtest[i])):
flag = 1
print(flag)
In [44]:
X_test["cat_0n"] = cat_0n_xtest
X_test["cat_1p"] = cat_1p_xtest
In [45]:
#https://www.analyticsvidhya.com/blog/2015/11/easy-methods-deal-categorical-variables-predictive-modeling/
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
cat_std0n = StandardScaler()
cat_std0n.fit(X_train['cat_0n'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
# Now standardize the data with above maen and variance.
cat_0n_xtrain = cat_std0n.transform(X_train['cat_0n'].values.reshape(-1, 1))
cat_0n_xtest = cat_std0n.transform(X_test['cat_0n'].values.reshape(-1, 1))
print(cat_0n_xtrain.shape)
print(cat_0n_xtest.shape)
In [46]:
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
cat_std1p = StandardScaler()
cat_std1p.fit(X_train['cat_1p'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
# Now standardize the data with above maen and variance.
cat_1p_xtrain = cat_std1p.transform(X_train['cat_1p'].values.reshape(-1, 1))
cat_1p_xtest = cat_std1p.transform(X_test['cat_1p'].values.reshape(-1, 1))
print(cat_1p_xtrain.shape)
print(cat_1p_xtest.shape)
2. project_subject_subcategories : clean_subcategories¶
In [47]:
clean_pos_sub_cat = {}
for a in Xtrain_pos['clean_subcategories'] :
for b in a.split():
if b not in clean_pos_sub_cat :
clean_pos_sub_cat[b] = 1
else :
clean_pos_sub_cat[b] += 1
clean_neg_sub_cat = {}
for a in Xtrain_neg['clean_subcategories'] :
for b in a.split():
if b not in clean_neg_sub_cat :
clean_neg_sub_cat[b] = 1
else :
clean_neg_sub_cat[b] += 1
clean_sub_cat_xtrain = {}
for a in X_train['clean_subcategories'] :
for b in a.split():
if b not in clean_sub_cat_xtrain :
clean_sub_cat_xtrain[b] = 1
else :
clean_sub_cat_xtrain[b] += 1
In [48]:
pos_sub_cat_p = {}
for p in clean_sub_cat_xtrain.keys():
pos_sub_cat_p[p] = (clean_pos_sub_cat[p])/float(clean_sub_cat_xtrain[p])
neg_sub_cat_n = {}
for n in clean_sub_cat_xtrain.keys():
neg_sub_cat_n[n] = (clean_neg_sub_cat[n])/float(clean_sub_cat_xtrain[n])
sub_cat_0n_xtrain = []
sub_cat_1p_xtrain = []
for a in X_train['clean_subcategories'] :
b = a.split()
if len(b) == 1 :
sub_cat_0n_xtrain.append(neg_sub_cat_n[a])
sub_cat_1p_xtrain.append(pos_sub_cat_p[a])
else :
c = neg_sub_cat_n[b[0]]
d =neg_sub_cat_n[b[1]]
e = pos_sub_cat_p[b[0]]
f = pos_sub_cat_p[b[1]]
sub_cat_0n_xtrain.append(c*d)
sub_cat_1p_xtrain.append(e*f)
X_train["sub_cat_0n"] = sub_cat_0n_xtrain
X_train["sub_cat_1p"] = sub_cat_1p_xtrain
In [49]:
import math
sub_cat_0n_xtest = []
sub_cat_1p_xtest = []
for a in X_test['clean_subcategories'] :
b = a.split()
if len(b) == 1 :
sub_cat_0n_xtest.append(neg_sub_cat_n[a])
sub_cat_1p_xtest.append(pos_sub_cat_p[a])
else :
c = neg_sub_cat_n[b[0]]
d = neg_sub_cat_n[b[1]]
e = pos_sub_cat_p[b[0]]
f = pos_sub_cat_p[b[1]]
sub_cat_0n_xtest.append(c*d)
sub_cat_1p_xtest.append(e*f)
flag = 0
for i in range(len(sub_cat_0n_xtest)):
if(math.isnan(sub_cat_0n_xtest[i])):
flag = 1
print(flag)
In [50]:
X_test["sub_cat_0n"] = cat_0n_xtest
X_test["sub_cat_1p"] = cat_1p_xtest
In [51]:
#https://www.analyticsvidhya.com/blog/2015/11/easy-methods-deal-categorical-variables-predictive-modeling/
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
sub_cat_std0n = StandardScaler()
sub_cat_std0n.fit(X_train['sub_cat_0n'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
# Now standardize the data with above maen and variance.
sub_cat_0n_xtrain = sub_cat_std0n.transform(X_train['sub_cat_0n'].values.reshape(-1, 1))
sub_cat_0n_xtest = cat_std0n.transform(X_test['sub_cat_0n'].values.reshape(-1, 1))
print(sub_cat_0n_xtrain.shape)
print(sub_cat_0n_xtest.shape)
In [52]:
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
sub_cat_std1p = StandardScaler()
sub_cat_std1p.fit(X_train['sub_cat_1p'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
# Now standardize the data with above maen and variance.
sub_cat_1p_xtrain = sub_cat_std1p.transform(X_train['sub_cat_1p'].values.reshape(-1, 1))
sub_cat_1p_xtest = sub_cat_std1p.transform(X_test['sub_cat_1p'].values.reshape(-1, 1))
print(sub_cat_1p_xtrain.shape)
print(sub_cat_1p_xtest.shape)
3. School State¶
In [53]:
school_state_pos = {}
for a in Xtrain_pos['clean_project_school_category'] :
if a not in school_state_pos :
school_state_pos[a] = 1
else :
school_state_pos[a] += 1
school_state_neg = {}
for a in Xtrain_neg['clean_project_school_category'] :
if a not in school_state_neg :
school_state_neg[a] = 1
else :
school_state_neg[a] += 1
school_state_xtrain = {}
for a in X_train['clean_project_school_category'] :
if a not in school_state_xtrain :
school_state_xtrain[a] = 1
else :
school_state_xtrain[a] += 1
In [54]:
pos_school_state_p = {}
for state in school_state_xtrain.keys():
pos_school_state_p[state] = (school_state_pos[state])/float(school_state_xtrain[state])
neg_school_state_n = {}
for state in school_state_xtrain.keys():
neg_school_state_n[state] = (school_state_neg[state])/float(school_state_xtrain[state])
school_state_0n_xtrain = []
school_state_1p_xtrain = []
for a in X_train['clean_project_school_category'] :
school_state_0n_xtrain.append(neg_school_state_n[a])
school_state_1p_xtrain.append(pos_school_state_p[a])
In [55]:
X_train["school_state_0n"] =school_state_0n_xtrain
X_train["school_state_1p"] = school_state_1p_xtrain
In [56]:
school_state_0n_xtest = []
school_state_1p_xtest = []
for a in X_test['clean_project_school_category'] :
school_state_0n_xtest.append(neg_school_state_n[a])
school_state_1p_xtest.append(pos_school_state_p[a])
flag = 0
for i in range(len(school_state_0n_xtest)):
if(math.isnan(school_state_0n_xtest[i])):
flag = 1
print(flag)
In [57]:
X_test["school_state_0n"] =school_state_0n_xtest
X_test["school_state_1p"] =school_state_1p_xtest
In [58]:
#https://www.analyticsvidhya.com/blog/2015/11/easy-methods-deal-categorical-variables-predictive-modeling/
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
school_state_std0 = StandardScaler()
school_state_std0.fit(X_train["school_state_0n"].values.reshape(-1,1)) # finding the mean and standard deviation of this data
school_state_0n_xtrain = school_state_std0.transform(X_train["school_state_0n"].values.reshape(-1, 1))
school_state_0n_xtest = school_state_std0.transform(X_test["school_state_0n"].values.reshape(-1, 1))
print(school_state_0n_xtrain.shape)
print(school_state_0n_xtest.shape)
In [59]:
#https://www.analyticsvidhya.com/blog/2015/11/easy-methods-deal-categorical-variables-predictive-modeling/
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
school_state_std1 = StandardScaler()
school_state_std1.fit(X_train["school_state_1p"].values.reshape(-1,1)) # finding the mean and standard deviation of this data
school_state_1p_xtrain = school_state_std1.transform(X_train["school_state_1p"].values.reshape(-1, 1))
school_state_1p_xtest = school_state_std1.transform(X_test["school_state_1p"].values.reshape(-1, 1))
print(school_state_1p_xtrain.shape)
print(school_state_1p_xtest.shape)
4. Teacher Prefix¶
In [60]:
teacher_prefix_pos = {}
for a in Xtrain_pos['clean_teacher_prefix_category'] :
if a not in teacher_prefix_pos :
teacher_prefix_pos[a] = 1
else :
teacher_prefix_pos[a] += 1
teacher_prefix_neg = {}
for a in Xtrain_neg['clean_teacher_prefix_category'] :
if a not in teacher_prefix_neg :
teacher_prefix_neg[a] = 1
else :
teacher_prefix_neg[a] += 1
teacher_prefix_xtrain = {}
for a in X_train['clean_teacher_prefix_category'] :
if a not in teacher_prefix_xtrain :
teacher_prefix_xtrain[a] = 1
else :
teacher_prefix_xtrain[a] += 1
In [61]:
pos_teacher_pref = {}
for p in teacher_prefix_xtrain.keys():
pos_teacher_pref[p] = (teacher_prefix_pos[p])/float(teacher_prefix_xtrain[p])
neg_teacher_pref = {}
for n in teacher_prefix_xtrain.keys():
neg_teacher_pref[n] = (teacher_prefix_neg[n])/float(teacher_prefix_xtrain[n])
teacher_pref_0n_xtrain = []
teacher_pref_1p_xtrain = []
for a in X_train['clean_teacher_prefix_category'] :
teacher_pref_0n_xtrain.append(neg_teacher_pref[a])
teacher_pref_1p_xtrain.append(pos_teacher_pref[a])
In [62]:
X_train["teacher_prefix_0n"] = teacher_pref_0n_xtrain
X_train["teacher_prefix_1p"] = teacher_pref_1p_xtrain
In [63]:
teacher_pref_0n_xtest = []
teacher_pref_1p_xtest = []
In [64]:
for a in X_test['clean_teacher_prefix_category'] :
teacher_pref_0n_xtest.append(neg_teacher_pref[a])
teacher_pref_1p_xtest.append(pos_teacher_pref[a])
In [65]:
flag = 0
for i in range(len(teacher_pref_0n_xtest)):
if(math.isnan(teacher_pref_0n_xtest[i])):
flag = 1
print(flag)
In [66]:
X_test["teacher_prefix_0n"] = teacher_pref_0n_xtest
X_test["teacher_prefix_1p"] = teacher_pref_1p_xtest
In [67]:
#https://www.analyticsvidhya.com/blog/2015/11/easy-methods-deal-categorical-variables-predictive-modeling/
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
teacher_pref_std0 = StandardScaler()
teacher_pref_std0.fit(X_train['teacher_prefix_0n'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
teacher_prefix_0n_xtrain = teacher_pref_std0.transform(X_train['teacher_prefix_0n'].values.reshape(-1, 1))
teacher_prefix_0n_xtest = teacher_pref_std0.transform(X_test['teacher_prefix_0n'].values.reshape(-1, 1))
print(teacher_prefix_0n_xtrain.shape)
print(teacher_prefix_0n_xtest.shape)
In [68]:
#https://www.analyticsvidhya.com/blog/2015/11/easy-methods-deal-categorical-variables-predictive-modeling/
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
teacher_pref_std1 = StandardScaler()
teacher_pref_std1.fit(X_train['teacher_prefix_1p'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
teacher_prefix_1p_xtrain = teacher_pref_std1.transform(X_train['teacher_prefix_1p'].values.reshape(-1, 1))
teacher_prefix_1p_xtest = teacher_pref_std1.transform(X_test['teacher_prefix_1p'].values.reshape(-1, 1))
print(teacher_prefix_1p_xtrain.shape)
print(teacher_prefix_1p_xtest.shape)
5. Project Grade¶
In [69]:
proj_grade_pos = {}
for a in Xtrain_pos['clean_project_grade_category'] :
if a not in proj_grade_pos :
proj_grade_pos[a] = 1
else :
proj_grade_pos[a] += 1
proj_grade_neg = {}
for a in Xtrain_neg['clean_project_grade_category'] :
if a not in proj_grade_neg :
proj_grade_neg[a] = 1
else :
proj_grade_neg[a] += 1
proj_grade_xtrain = {}
for a in X_train['clean_project_grade_category'] :
if a not in proj_grade_xtrain :
proj_grade_xtrain[a] = 1
else :
proj_grade_xtrain[a] += 1
In [70]:
pos_proj_grade = {}
for p in proj_grade_xtrain.keys():
pos_proj_grade[p] = (proj_grade_pos[p])/float(proj_grade_xtrain[p])
neg_proj_grade = {}
for n in proj_grade_xtrain.keys():
neg_proj_grade[n] = (proj_grade_neg[n])/float(proj_grade_xtrain[n])
proj_grade_0n_xtrain = []
proj_grade_1p_xtrain = []
for a in X_train["clean_project_grade_category"] :
proj_grade_0n_xtrain.append(neg_proj_grade[a])
proj_grade_1p_xtrain.append(pos_proj_grade[a])
In [71]:
X_train["proj_grade_0n"] = proj_grade_0n_xtrain
X_train["proj_grade_1p"] = proj_grade_1p_xtrain
In [72]:
proj_grade_0n_xtest = []
proj_grade_1p_xtest = []
for a in X_test["clean_project_grade_category"] :
proj_grade_0n_xtest.append(neg_proj_grade[a])
proj_grade_1p_xtest.append(pos_proj_grade[a])
flag = 0
for i in range(len(proj_grade_0n_xtest)):
if(math.isnan(proj_grade_0n_xtest[i])):
flag = 1
print(flag)
In [73]:
X_test["proj_grade_0n"] = proj_grade_0n_xtest
X_test["proj_grade_1p"] = proj_grade_1p_xtest
In [74]:
#https://www.analyticsvidhya.com/blog/2015/11/easy-methods-deal-categorical-variables-predictive-modeling/
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
proj_grade_std0 = StandardScaler()
proj_grade_std0.fit(X_train['proj_grade_0n'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
proj_grade_0n_xtrain = proj_grade_std0 .transform(X_train['proj_grade_0n'].values.reshape(-1, 1))
proj_grade_0n_xtest = proj_grade_std0 .transform(X_test['proj_grade_0n'].values.reshape(-1, 1))
print(proj_grade_0n_xtrain.shape)
print(proj_grade_0n_xtest.shape)
In [76]:
#https://www.analyticsvidhya.com/blog/2015/11/easy-methods-deal-categorical-variables-predictive-modeling/
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
proj_grade_std1 = StandardScaler()
proj_grade_std1.fit(X_train['proj_grade_1p'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
proj_grade_1p_xtrain = proj_grade_std1.transform(X_train['proj_grade_1p'].values.reshape(-1, 1))
proj_grade_1p_xtest = proj_grade_std1.transform(X_test['proj_grade_1p'].values.reshape(-1, 1))
print(proj_grade_1p_xtrain.shape)
print(proj_grade_1p_xtest.shape)
Vectorizing Numerical features¶
In [77]:
# check this one: https://www.youtube.com/watch?v=0HOqOcln3Z4&t=530s
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
# price_standardized = standardScalar.fit(project_data['price'].values)
# this will rise the error
# ValueError: Expected 2D array, got 1D array instead: array=[725.05 213.03 329. ... 399. 287.73 5.5 ].
# Reshape your data either using array.reshape(-1, 1)
price_scalar = StandardScaler()
price_scalar.fit(X_train['price'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
# print(f"Mean : {price_scalar.mean_[0]}, Standard deviation : {np.sqrt(price_scalar.var_[0])}")
# Now standardize the data with above maen and variance.
price_standardized_xtrain = price_scalar.transform(X_train['price'].values.reshape(-1, 1))
#price_standardized_xcv = price_scalar.transform(X_cv['price'].values.reshape(-1, 1))
price_standardized_xtest = price_scalar.transform(X_test['price'].values.reshape(-1, 1))
print("shape of price_standardized_xtrain",price_standardized_xtrain.shape)
#print("shape of price_standardized_xcv",price_standardized_xcv.shape)
print("shape of price_standardized_xtest",price_standardized_xtest.shape)
In [78]:
# check this one: https://www.youtube.com/watch?v=0HOqOcln3Z4&t=530s
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
# price_standardized = standardScalar.fit(project_data['price'].values)
# this will rise the error
# ValueError: Expected 2D array, got 1D array instead: array=[725.05 213.03 329. ... 399. 287.73 5.5 ].
# Reshape your data either using array.reshape(-1, 1)
quantity_scalar = StandardScaler()
quantity_scalar.fit(X_train['quantity'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
# print(f"Mean : {quantity_scalar.mean_[0]}, Standard deviation : {np.sqrt(quantity_scalar.var_[0])}")
# Now standardize the data with above maen and variance.
quantity_standardized_xtrain = quantity_scalar.transform(X_train['quantity'].values.reshape(-1, 1))
#quantity_standardized_xcv = quantity_scalar.transform(X_cv['quantity'].values.reshape(-1, 1))
quantity_standardized_xtest = quantity_scalar.transform(X_test['quantity'].values.reshape(-1, 1))
print("shape of quantity_standardized_xtrain",quantity_standardized_xtrain.shape)
#print("shape of quantity_standardized_xcv",quantity_standardized_xcv.shape)
print("shape of quantity_standardized_xtest",quantity_standardized_xtest.shape)
In [79]:
# check this one: https://www.youtube.com/watch?v=0HOqOcln3Z4&t=530s
# standardization sklearn: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
from sklearn.preprocessing import StandardScaler
# price_standardized = standardScalar.fit(project_data['price'].values)
# this will rise the error
# ValueError: Expected 2D array, got 1D array instead: array=[725.05 213.03 329. ... 399. 287.73 5.5 ].
# Reshape your data either using array.reshape(-1, 1)
teacher_num_prev_projects_scalar = StandardScaler()
teacher_num_prev_projects_scalar.fit(X_train['teacher_number_of_previously_posted_projects'].values.reshape(-1,1)) # finding the mean and standard deviation of this data
# print(f"Mean : {teacher_number_of_previously_posted_projects_scalar.mean_[0]}, Standard deviation : {np.sqrt(teacher_number_of_previously_posted_projects_scalar.var_[0])}")
# Now standardize the data with above maen and variance.
teacher_num_prev_projects_standardized_xtrain = teacher_num_prev_projects_scalar.transform(X_train['teacher_number_of_previously_posted_projects'].values.reshape(-1, 1))
#teacher_num_prev_projects_standardized_xcv = teacher_num_prev_projects_scalar.transform(X_cv['teacher_number_of_previously_posted_projects'].values.reshape(-1, 1))
teacher_num_prev_projects_standardized_xtest = teacher_num_prev_projects_scalar.transform(X_test['teacher_number_of_previously_posted_projects'].values.reshape(-1, 1))
print(" shape of teacher_number_of_previously_posted_projects_standardized_xtrain",teacher_num_prev_projects_standardized_xtrain.shape)
#print(" shape of teacher_number_of_previously_posted_projects_standardized_xcv",teacher_num_prev_projects_standardized_xcv.shape)
print(" shape of teacher_number_of_previously_posted_projects_standardized_xtest",teacher_num_prev_projects_standardized_xtest.shape)
2.3 Make Data Model Ready: encoding eassay, and project_title
In [ ]:
# please write all the code with proper documentation, and proper titles for each subsection
# go through documentations and blogs before you start coding
# first figure out what to do, and then think about how to do.
# reading and understanding error messages will be very much helpfull in debugging your code
# make sure you featurize train and test data separatly
# when you plot any graph make sure you use
# a. Title, that describes your plot, this will be very helpful to the reader
# b. Legends if needed
# c. X-axis label
# d. Y-axis label
In [80]:
# BOW on eassay
# We are considering only the words which appeared in at least 10 documents(rows or projects).
vectorizer_bow_essays = CountVectorizer(min_df=10,max_features=5000,ngram_range=(1,2))
vectorizer_bow_essays.fit(X_train['preprocessed_essays'])
essay_text_bow_xtrain = vectorizer_bow_essays.transform(X_train['preprocessed_essays'])
#essay_text_bow_xcv = vectorizer_bow_essays.transform(X_cv['preprocessed_essays'])
essay_text_bow_xtest = vectorizer_bow_essays.transform(X_test['preprocessed_essays'])
print("Shape of matrix after BOW_text_essay X_train ",essay_text_bow_xtrain.shape)
#print("Shape of matrix after BOW_text_essay X_cv ",essay_text_bow_xcv.shape)
print("Shape of matrix after BOW_text_essay X_test ",essay_text_bow_xtest.shape)
BOW on project_title¶
In [81]:
# BOW on project_title
# We are considering only the words which appeared in at least 10 documents(rows or projects).
vectorizer_bow_titles = CountVectorizer(min_df=10)
vectorizer_bow_titles.fit(X_train['preprocessed_titles'])
proj_title_bow_xtrain = vectorizer_bow_titles.transform(X_train['preprocessed_titles'])
#proj_title_bow_xcv = vectorizer_bow_titles.transform(X_cv['preprocessed_titles'])
proj_title_bow_xtest = vectorizer_bow_titles.transform(X_test['preprocessed_titles'])
print("Shape of matrix after BOW project_title_xtrain ",proj_title_bow_xtrain.shape)
#print("Shape of matrix after BOW project_title_xcv ",proj_title_bow_xcv.shape)
print("Shape of matrix after BOW project_title_xtest ",proj_title_bow_xtest.shape)
TFIDF Vectorizer on Essay¶
In [82]:
from sklearn.feature_extraction.text import TfidfVectorizer
vectorizer_tfidf_essays = TfidfVectorizer(min_df=10,max_features=5000,ngram_range=(1,2))
vectorizer_tfidf_essays.fit(X_train['preprocessed_essays'])
essay_tfidf_xtrain = vectorizer_tfidf_essays.transform(X_train['preprocessed_essays'])
#essay_tfidf_xcv = vectorizer_tfidf_essays.transform(X_cv['preprocessed_essays'])
essay_tfidf_xtest = vectorizer_tfidf_essays.transform(X_test['preprocessed_essays'])
print("Shape of matrix after tfidf eassay_xtrain ",essay_tfidf_xtrain.shape)
#print("Shape of matrix after tfidf essay_xcv ",essay_tfidf_xcv.shape)
print("Shape of matrix after tfidf essay_xtest ",essay_tfidf_xtest.shape)
TFIDF Vectorizer on Project Title¶
In [83]:
from sklearn.feature_extraction.text import TfidfVectorizer
vectorizer_tfidf_title = TfidfVectorizer(min_df=10)
vectorizer_tfidf_title.fit(X_train['preprocessed_titles'])
proj_title_tfidf_xtrain = vectorizer_tfidf_title.transform(X_train['preprocessed_titles'])
#proj_title_tfidf_xcv = vectorizer_tfidf_title.transform(X_cv['preprocessed_titles'])
proj_title_tfidf_xtest = vectorizer_tfidf_title.transform(X_test['preprocessed_titles'])
print("Shape of matrix after tfidf proj_title_xtrain ",proj_title_tfidf_xtrain.shape)
#print("Shape of matrix after tfidf proj_title_xcv ",proj_title_tfidf_xcv.shape)
print("Shape of matrix after tfidf proj_title_xtest ",proj_title_tfidf_xtest.shape)
In [84]:
# Using Pretrained Models: Avg W2V
In [85]:
# stronging variables into pickle files python: http://www.jessicayung.com/how-to-use-pickle-to-save-and-load-variables-in-python/
# make sure you have the glove_vectors file
with open('glove_vectors', 'rb') as f:
model = pickle.load(f)
glove_words = set(model.keys())
Average Word2Vec on Essay¶
In [86]:
# average Word2Vec
# compute average word2vec for each review.
# average Word2Vec on X_train
essay_avg_w2v_vectors_xtrain = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(X_train['preprocessed_essays']): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length
cnt_words =0; # num of words with a valid vector in the sentence/review
for word in sentence.split(): # for each word in a review/sentence
if word in glove_words:
vector += model[word]
cnt_words += 1
if cnt_words != 0:
vector /= cnt_words
essay_avg_w2v_vectors_xtrain.append(vector)
print(len(essay_avg_w2v_vectors_xtrain))
print(len(essay_avg_w2v_vectors_xtrain[0]))
# average Word2Vec on X_test
essay_avg_w2v_vectors_xtest = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(X_test['preprocessed_essays']): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length
cnt_words =0; # num of words with a valid vector in the sentence/review
for word in sentence.split(): # for each word in a review/sentence
if word in glove_words:
vector += model[word]
cnt_words += 1
if cnt_words != 0:
vector /= cnt_words
essay_avg_w2v_vectors_xtest.append(vector)
print(len(essay_avg_w2v_vectors_xtest))
print(len(essay_avg_w2v_vectors_xtest[0]))
Average Word2Vec on Project Title¶
In [87]:
# average Word2Vec
# compute average word2vec for each review.
# average Word2Vec on X_train
proj_title_avg_w2v_vectors_xtrain = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(X_train['preprocessed_titles']): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length
cnt_words =0; # num of words with a valid vector in the sentence/review
for word in sentence.split(): # for each word in a review/sentence
if word in glove_words:
vector += model[word]
cnt_words += 1
if cnt_words != 0:
vector /= cnt_words
proj_title_avg_w2v_vectors_xtrain.append(vector)
print(len(proj_title_avg_w2v_vectors_xtrain))
print(len(proj_title_avg_w2v_vectors_xtrain[0]))
# average Word2Vec on X_test
proj_title_avg_w2v_vectors_xtest = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(X_test['preprocessed_titles']): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length
cnt_words =0; # num of words with a valid vector in the sentence/review
for word in sentence.split(): # for each word in a review/sentence
if word in glove_words:
vector += model[word]
cnt_words += 1
if cnt_words != 0:
vector /= cnt_words
proj_title_avg_w2v_vectors_xtest.append(vector)
print(len(proj_title_avg_w2v_vectors_xtest))
print(len(proj_title_avg_w2v_vectors_xtest[0]))
In [88]:
# S = ["abc def pqr", "def def def abc", "pqr pqr def"]
tfidf_model = TfidfVectorizer()
tfidf_model.fit(X_train['preprocessed_essays'])
# we are converting a dictionary with word as a key, and the idf as a value
dictionary = dict(zip(tfidf_model.get_feature_names(), list(tfidf_model.idf_)))
tfidf_words = set(tfidf_model.get_feature_names())
In [89]:
# average Word2Vec
# average Word2Vec on X_train
essay_tfidf_w2v_vectors_xtrain = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(X_train['preprocessed_essays']): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length
tf_idf_weight =0; # num of words with a valid vector in the sentence/review
for word in sentence.split(): # for each word in a review/sentence
if (word in glove_words) and (word in tfidf_words):
vec = model[word] # getting the vector for each word
# here we are multiplying idf value(dictionary[word]) and the tf value((sentence.count(word)/len(sentence.split())))
tf_idf = dictionary[word]*(sentence.count(word)/len(sentence.split())) # getting the tfidf value for each word
vector += (vec * tf_idf) # calculating tfidf weighted w2v
tf_idf_weight += tf_idf
if tf_idf_weight != 0:
vector /= tf_idf_weight
essay_tfidf_w2v_vectors_xtrain.append(vector)
print(len(essay_tfidf_w2v_vectors_xtrain))
print(len(essay_tfidf_w2v_vectors_xtrain[0]))
# average Word2Vec on X_train
essay_tfidf_w2v_vectors_xtest = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(X_test['preprocessed_essays']): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length
tf_idf_weight =0; # num of words with a valid vector in the sentence/review
for word in sentence.split(): # for each word in a review/sentence
if (word in glove_words) and (word in tfidf_words):
vec = model[word] # getting the vector for each word
# here we are multiplying idf value(dictionary[word]) and the tf value((sentence.count(word)/len(sentence.split())))
tf_idf = dictionary[word]*(sentence.count(word)/len(sentence.split())) # getting the tfidf value for each word
vector += (vec * tf_idf) # calculating tfidf weighted w2v
tf_idf_weight += tf_idf
if tf_idf_weight != 0:
vector /= tf_idf_weight
essay_tfidf_w2v_vectors_xtest.append(vector)
print(len(essay_tfidf_w2v_vectors_xtest))
print(len(essay_tfidf_w2v_vectors_xtest[0]))
TFIDF weighted W2V on Project Title¶
In [90]:
# S = ["abc def pqr", "def def def abc", "pqr pqr def"]
tfidf_model = TfidfVectorizer()
tfidf_model.fit(X_train['preprocessed_titles'])
# we are converting a dictionary with word as a key, and the idf as a value
dictionary = dict(zip(tfidf_model.get_feature_names(), list(tfidf_model.idf_)))
tfidf_words = set(tfidf_model.get_feature_names())
In [91]:
# TFIDF weighted W2V on Project Title
# compute average word2vec for each review.
# TFIDF weighted W2V on X_train
proj_title_tfidf_w2v_vectors_xtrain = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(X_train['preprocessed_titles']): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length
tf_idf_weight =0; # num of words with a valid vector in the sentence/review
for word in sentence.split(): # for each word in a review/sentence
if (word in glove_words) and (word in tfidf_words):
vec = model[word] # getting the vector for each word
# here we are multiplying idf value(dictionary[word]) and the tf value((sentence.count(word)/len(sentence.split())))
tf_idf = dictionary[word]*(sentence.count(word)/len(sentence.split())) # getting the tfidf value for each word
vector += (vec * tf_idf) # calculating tfidf weighted w2v
tf_idf_weight += tf_idf
if tf_idf_weight != 0:
vector /= tf_idf_weight
proj_title_tfidf_w2v_vectors_xtrain.append(vector)
print(len(proj_title_tfidf_w2v_vectors_xtrain))
print(len(proj_title_tfidf_w2v_vectors_xtrain[0]))
# TFIDF weighted W2V on X_test
proj_title_tfidf_w2v_vectors_xtest = []; # the avg-w2v for each sentence/review is stored in this list
for sentence in tqdm(X_test['preprocessed_titles']): # for each review/sentence
vector = np.zeros(300) # as word vectors are of zero length
tf_idf_weight =0; # num of words with a valid vector in the sentence/review
for word in sentence.split(): # for each word in a review/sentence
if (word in glove_words) and (word in tfidf_words):
vec = model[word] # getting the vector for each word
# here we are multiplying idf value(dictionary[word]) and the tf value((sentence.count(word)/len(sentence.split())))
tf_idf = dictionary[word]*(sentence.count(word)/len(sentence.split())) # getting the tfidf value for each word
vector += (vec * tf_idf) # calculating tfidf weighted w2v
tf_idf_weight += tf_idf
if tf_idf_weight != 0:
vector /= tf_idf_weight
proj_title_tfidf_w2v_vectors_xtest.append(vector)
print(len(proj_title_tfidf_w2v_vectors_xtest))
print(len(proj_title_tfidf_w2v_vectors_xtest[0]))
2.4 Applying Random Forest
In [92]:
# please write all the code with proper documentation, and proper titles for each subsection
# go through documentations and blogs before you start coding
# first figure out what to do, and then think about how to do.
# reading and understanding error messages will be very much helpfull in debugging your code
# when you plot any graph make sure you use
# a. Title, that describes your plot, this will be very helpful to the reader
# b. Legends if needed
# c. X-axis label
# d. Y-axis label
2.4.1 Applying Random Forest on BOW, SET 1¶
In [93]:
# Please write all the code with proper documentation
# merge two sparse matrices: https://stackoverflow.com/a/19710648/4084039
from scipy.sparse import hstack
X_train1=hstack((cat_0n_xtrain, cat_1p_xtrain, sub_cat_0n_xtrain, sub_cat_1p_xtrain, school_state_0n_xtrain,
school_state_1p_xtrain, teacher_prefix_0n_xtrain, teacher_prefix_1p_xtrain,
proj_grade_0n_xtrain, proj_grade_1p_xtrain, price_standardized_xtrain,
teacher_num_prev_projects_standardized_xtrain, quantity_standardized_xtrain,
essay_text_bow_xtrain, proj_title_bow_xtrain)).tocsr()
X_test1=hstack((cat_0n_xtest, cat_1p_xtest, sub_cat_0n_xtest, sub_cat_1p_xtest, school_state_0n_xtest,
school_state_1p_xtest, teacher_prefix_0n_xtest, teacher_prefix_1p_xtest,
proj_grade_0n_xtest, proj_grade_1p_xtest, price_standardized_xtest,
teacher_num_prev_projects_standardized_xtest, quantity_standardized_xtest,
essay_text_bow_xtest, proj_title_bow_xtest)).tocsr()
print(X_train1.shape, y_train.shape)
print(X_test1.shape, y_test.shape)
GridSearchCV¶
In [94]:
import warnings
warnings.filterwarnings('ignore')
import time
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import GridSearchCV
start_time = time.time()
rfclf1 = RandomForestClassifier(n_jobs=-1,class_weight='balanced')
parameters = {'n_estimators': [10, 300, 500, 700], 'max_depth':[10, 30, 60, 100]}
clf1 = GridSearchCV(rfclf1, parameters, cv= 5, scoring='roc_auc',return_train_score=True)
clf1.fit(X_train1, y_train)
train_auc= clf1.cv_results_['mean_train_score']
train_auc_std= clf1.cv_results_['std_train_score']
cv_auc = clf1.cv_results_['mean_test_score']
cv_auc_std= clf1.cv_results_['std_test_score']
print("Total Execution time: " + str((time.time() - start_time)) + ' ms')
In [ ]:
## 8.15-11.17=3 hours
In [95]:
train_auc = train_auc.reshape(4,4)
train_auc
Out[95]:
In [96]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(train_auc,annot=True)
plt.xticks(np.arange(4), [10, 30, 60, 100])
plt.yticks(np.arange(4), [10, 300, 500, 700])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [97]:
cv_auc = cv_auc.reshape(4,4)
cv_auc
Out[97]:
In [98]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(cv_auc,annot=True)
plt.xticks(np.arange(4), [10, 30, 60, 100])
plt.yticks(np.arange(4), [10, 300, 500, 700])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
Simple for loop (if you are having memory limitations use this)¶
In [100]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [107]:
from sklearn.metrics import roc_curve, auc
modelbow = RandomForestClassifier(max_depth = 10, n_estimators = 500,n_jobs=-1,class_weight='balanced')
modelbow.fit(X_train1, y_train)
y_train_pred = batch_predict(modelbow, X_train1)
y_test_pred = batch_predict(modelbow, X_test1)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(FPR)")
plt.ylabel("True Positive Rate(TPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [108]:
# we are writing our own function for predict, with defined thresould
# we will pick a threshold that will give the least fpr
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(tpr*(1-fpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
return predictions
In [109]:
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train, predict(y_train_pred, tr_thresholds, train_fpr, train_fpr)))
print("Test confusion matrix")
print(confusion_matrix(y_test, predict(y_test_pred, tr_thresholds, test_fpr, test_fpr)))
In [110]:
# Confusion matrix for train data
# Code for this segment from here -->> https://stackoverflow.com/questions/35572000/how-can-i-plot-a-confusion-matrix
conf_matrix_xtrain = pd.DataFrame(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matrix_xtrain, annot=True,annot_kws={"size": 16}, fmt='g')# font size
Out[110]:
In [111]:
# Confusion matrix for test data
# Code for this segment from here -->> https://stackoverflow.com/questions/35572000/how-can-i-plot-a-confusion-matrix
conf_matrix_xtest = pd.DataFrame(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matrix_xtest, annot=True,annot_kws={"size": 16}, fmt='g')#font size
Out[111]:
In [1]:
import dill
#dill.dump_session('notebook_4_11.db')
dill.load_session('notebook_4_11.db')
2.4.2 Applying Logistic Regression on TFIDF, SET 2¶
In [2]:
# Please write all the code with proper documentation
# merge two sparse matrices: https://stackoverflow.com/a/19710648/4084039
from scipy.sparse import hstack
X_train2=hstack((cat_0n_xtrain, cat_1p_xtrain, sub_cat_0n_xtrain, sub_cat_1p_xtrain, school_state_0n_xtrain,
school_state_1p_xtrain, teacher_prefix_0n_xtrain, teacher_prefix_1p_xtrain,
proj_grade_0n_xtrain, proj_grade_1p_xtrain, price_standardized_xtrain,
teacher_num_prev_projects_standardized_xtrain,
quantity_standardized_xtrain,essay_tfidf_xtrain, proj_title_tfidf_xtrain)).tocsr()
X_test2=hstack((cat_0n_xtest, cat_1p_xtest, sub_cat_0n_xtest, sub_cat_1p_xtest, school_state_0n_xtest,
school_state_1p_xtest, teacher_prefix_0n_xtest, teacher_prefix_1p_xtest,
proj_grade_0n_xtest, proj_grade_1p_xtest, price_standardized_xtest,
teacher_num_prev_projects_standardized_xtest, quantity_standardized_xtest,
essay_tfidf_xtest, proj_title_tfidf_xtest)).tocsr()
print(X_train2.shape)
print(X_test2.shape)
GridSearchCV¶
In [3]:
import warnings
warnings.filterwarnings('ignore')
import time
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import GridSearchCV
start_time = time.time()
rfclf2 = RandomForestClassifier(n_jobs=-1,class_weight='balanced')
parameters = {'n_estimators': [10, 300, 500, 700], 'max_depth':[10, 30, 60, 100]}
clf2 = GridSearchCV(rfclf2, parameters, cv= 5, scoring='roc_auc',return_train_score=True)
clf2.fit(X_train2, y_train)
train_auc= clf2.cv_results_['mean_train_score']
train_auc_std= clf2.cv_results_['std_train_score']
cv_auc = clf2.cv_results_['mean_test_score']
cv_auc_std= clf2.cv_results_['std_test_score']
print("Total Execution time: " + str((time.time() - start_time)) + ' ms')
In [1]:
import dill
#dill.dump_session('notebook_44_11.db')
dill.load_session('notebook_44_11.db')
In [2]:
train_auc = train_auc.reshape(4,4)
train_auc
Out[2]:
In [3]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(train_auc,annot=True)
plt.xticks(np.arange(4), [10, 30, 60, 100])
plt.yticks(np.arange(4), [10, 300, 500, 700])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [4]:
cv_auc =cv_auc.reshape(4,4)
cv_auc
Out[4]:
In [5]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(cv_auc,annot=True)
plt.xticks(np.arange(4), [10, 30, 60, 100])
plt.yticks(np.arange(4), [10, 300, 500, 700])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
Simple for loop (if you are having memory limitations use this)¶
In [7]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [16]:
# https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve
from sklearn.metrics import roc_curve, auc
modeltfidf = RandomForestClassifier(max_depth = 10, n_estimators = 100,n_jobs=-1,class_weight='balanced')
modeltfidf.fit(X_train2, y_train)
y_train_pred = batch_predict(modeltfidf, X_train2)
y_test_pred = batch_predict(modeltfidf, X_test2)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(FPR)")
plt.ylabel("True Positive Rate(TPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [17]:
# we are writing our own function for predict, with defined thresould
# we will pick a threshold that will give the least fpr
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(tpr*(1-fpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
return predictions
In [18]:
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
print("Test confusion matrix")
print(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
In [19]:
# Confusion Matrix for Train Data
# Code for this segment from here -->> https://stackoverflow.com/questions/35572000/how-can-i-plot-a-confusion-matrix
conf_matrix_xtrain = pd.DataFrame(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matrix_xtrain, annot=True,annot_kws={"size": 16}, fmt='g')#font size
Out[19]:
In [20]:
# Confusion matrix for test data
# Code for this segment from here -->> https://stackoverflow.com/questions/35572000/how-can-i-plot-a-confusion-matrix
conf_matrix_xtest = pd.DataFrame(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matrix_xtest, annot=True,annot_kws={"size": 16}, fmt='g')#font size
Out[20]:
2.4.1 Applying Logistic Regression on AVG W2V, SET 3¶
In [22]:
essay_avg_w2v_vectors_xtrain= np.array(essay_avg_w2v_vectors_xtrain)
proj_title_avg_w2v_vectors_xtrain= np.array(proj_title_avg_w2v_vectors_xtrain)
In [23]:
essay_avg_w2v_vectors_xtest= np.array(essay_avg_w2v_vectors_xtest)
proj_title_avg_w2v_vectors_xtest= np.array(proj_title_avg_w2v_vectors_xtest)
In [26]:
# Please write all the code with proper documentation
# merge two sparse matrices: https://stackoverflow.com/a/19710648/4084039
from scipy.sparse import hstack
X_train3=np.hstack((cat_0n_xtrain, cat_1p_xtrain, sub_cat_0n_xtrain, sub_cat_1p_xtrain, school_state_0n_xtrain,
school_state_1p_xtrain, teacher_prefix_0n_xtrain, teacher_prefix_1p_xtrain,
proj_grade_0n_xtrain, proj_grade_1p_xtrain, price_standardized_xtrain,
teacher_num_prev_projects_standardized_xtrain,
essay_avg_w2v_vectors_xtrain, proj_title_avg_w2v_vectors_xtrain))
X_test3=np.hstack((cat_0n_xtest, cat_1p_xtest, sub_cat_0n_xtest, sub_cat_1p_xtest, school_state_0n_xtest,
school_state_1p_xtest, teacher_prefix_0n_xtest, teacher_prefix_1p_xtest,
proj_grade_0n_xtest, proj_grade_1p_xtest, price_standardized_xtest,
teacher_num_prev_projects_standardized_xtest,
essay_avg_w2v_vectors_xtest, proj_title_avg_w2v_vectors_xtest))
print(X_train3.shape, y_train.shape)
print(X_test3.shape, y_test.shape)
GridSearchCV¶
In [27]:
import warnings
warnings.filterwarnings('ignore')
import time
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import GridSearchCV
start_time = time.time()
rfclf3 = RandomForestClassifier(n_jobs=-1,class_weight='balanced')
parameters = {'n_estimators': [10, 300, 500, 700], 'max_depth':[10, 30, 60, 100]}
clf3 = GridSearchCV(rfclf3, parameters, cv= 5, scoring='roc_auc',return_train_score=True)
clf3.fit(X_train3, y_train)
train_auc= clf3.cv_results_['mean_train_score']
train_auc_std= clf3.cv_results_['std_train_score']
cv_auc = clf3.cv_results_['mean_test_score']
cv_auc_std= clf3.cv_results_['std_test_score']
print("Total Execution time: " + str((time.time() - start_time)) + ' ms')
In [ ]:
# 9.53am-5.28pm
In [30]:
train_auc = train_auc.reshape(4,4)
train_auc
Out[30]:
In [31]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(train_auc,annot=True)
plt.xticks(np.arange(4), [10, 30, 60, 100])
plt.yticks(np.arange(4), [10, 300, 500, 700])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [32]:
cv_auc = train_auc.reshape(4,4)
cv_auc
Out[32]:
In [33]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(cv_auc,annot=True)
plt.xticks(np.arange(4), [10, 30, 60, 100])
plt.yticks(np.arange(4), [10, 300, 500, 700])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
Simple for loop (if you are having memory limitations use this)¶
In [34]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [44]:
# https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve
from sklearn.metrics import roc_curve, auc
modelavgw2v = RandomForestClassifier(max_depth = 10, n_estimators = 10,n_jobs=-1,class_weight='balanced')
modelavgw2v.fit(X_train3, y_train)
y_train_pred = batch_predict(modelavgw2v, X_train3)
y_test_pred = batch_predict(modelavgw2v, X_test3)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(FPR)")
plt.ylabel("True Positive Rate(TPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [45]:
# we are writing our own function for predict, with defined thresould
# we will pick a threshold that will give the least fpr
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(tpr*(1-fpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
return predictions
In [46]:
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
print("Test confusion matrix")
print(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
In [47]:
# Confusion Matrix for Train Data
# Code for this segment from here -->> https://stackoverflow.com/questions/35572000/how-can-i-plot-a-confusion-matrix
conf_matrix_xtrain = pd.DataFrame(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matrix_xtrain, annot=True,annot_kws={"size": 16}, fmt='g')
Out[47]:
In [48]:
# Confusion matrix for test data
# Code for this segment from here -->> https://stackoverflow.com/questions/35572000/how-can-i-plot-a-confusion-matrix
conf_matrix_xtest = pd.DataFrame(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matrix_xtest, annot=True,annot_kws={"size": 16}, fmt='g')
Out[48]:
2.4.1 Applying Logistic Regression on TFIDF Word2Vec, SET 4¶
In [49]:
essay_tfidf_w2v_vectors_xtrain=np.array(essay_tfidf_w2v_vectors_xtrain)
proj_title_tfidf_w2v_vectors_xtrain=np.array(proj_title_tfidf_w2v_vectors_xtrain)
In [50]:
essay_tfidf_w2v_vectors_xtest=np.array(essay_tfidf_w2v_vectors_xtest)
proj_title_tfidf_w2v_vectors_xtest=np.array(proj_title_tfidf_w2v_vectors_xtest)
In [51]:
# Please write all the code with proper documentation
# merge two sparse matrices: https://stackoverflow.com/a/19710648/4084039
from scipy.sparse import hstack
X_train4=np.hstack((cat_0n_xtrain, cat_1p_xtrain, sub_cat_0n_xtrain, sub_cat_1p_xtrain, school_state_0n_xtrain,
school_state_1p_xtrain, teacher_prefix_0n_xtrain, teacher_prefix_1p_xtrain,
proj_grade_0n_xtrain, proj_grade_1p_xtrain, price_standardized_xtrain,
teacher_num_prev_projects_standardized_xtrain, quantity_standardized_xtrain,
essay_tfidf_w2v_vectors_xtrain, proj_title_tfidf_w2v_vectors_xtrain))
X_test4=np.hstack((cat_0n_xtest, cat_1p_xtest, sub_cat_0n_xtest, sub_cat_1p_xtest, school_state_0n_xtest,
school_state_1p_xtest, teacher_prefix_0n_xtest, teacher_prefix_1p_xtest,
proj_grade_0n_xtest, proj_grade_1p_xtest, price_standardized_xtest,
teacher_num_prev_projects_standardized_xtest, quantity_standardized_xtest,
essay_tfidf_w2v_vectors_xtest, proj_title_tfidf_w2v_vectors_xtest))
print(X_train4.shape, y_train.shape)
print(X_test4.shape, y_test.shape)
GridSearchCV¶
In [52]:
import warnings
warnings.filterwarnings('ignore')
import time
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import GridSearchCV
start_time = time.time()
rfclf4 = RandomForestClassifier(n_jobs=-1,class_weight='balanced')
parameters = {'n_estimators': [5, 10, 50, 100], 'max_depth':[2, 5, 7, 10]}
clf4 = GridSearchCV(rfclf4, parameters, cv= 5, scoring='roc_auc',return_train_score=True)
clf4.fit(X_train4, y_train)
train_auc= clf4.cv_results_['mean_train_score']
train_auc_std= clf4.cv_results_['std_train_score']
cv_auc = clf4.cv_results_['mean_test_score']
cv_auc_std= clf4.cv_results_['std_test_score']
print("Total Execution time: " + str((time.time() - start_time)) + ' ms')
In [ ]:
#6.36pm-7.05
In [53]:
# Testing the performance of the model on test data, plotting ROC Curves
# Select best log(C) value
best_set_tfidfw2v = clf4.best_params_
print(best_set_tfidfw2v)
In [54]:
train_auc = train_auc.reshape(4,4)
train_auc
Out[54]:
In [55]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(train_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [56]:
cv_auc = cv_auc.reshape(4,4)
cv_auc
Out[56]:
In [57]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(cv_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [58]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [59]:
# https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_curve.html#sklearn.metrics.roc_curve
from sklearn.metrics import roc_curve, auc
modeltfidfw2v = RandomForestClassifier(max_depth = 7, n_estimators = 100,n_jobs=-1,class_weight='balanced')
modeltfidfw2v.fit(X_train4, y_train)
y_train_pred = batch_predict(modeltfidfw2v, X_train4)
y_test_pred = batch_predict(modeltfidfw2v, X_test4)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(FPR)")
plt.ylabel("True Positive Rate(TPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [60]:
# we are writing our own function for predict, with defined thresould
# we will pick a threshold that will give the least fpr
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(tpr*(1-fpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
return predictions
In [61]:
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
print("Test confusion matrix")
print(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
In [62]:
# Confusion Matrix for Train Data
# Code for this segment from here -->> https://stackoverflow.com/questions/35572000/how-can-i-plot-a-confusion-matrix
conf_matrix_xtrain = pd.DataFrame(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matrix_xtrain, annot=True,annot_kws={"size": 16}, fmt='g')
Out[62]:
In [63]:
# Confusion matrix for test data
# Code for this segment from here -->> https://stackoverflow.com/questions/35572000/how-can-i-plot-a-confusion-matrix
conf_matrix_xtest = pd.DataFrame(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matrix_xtest, annot=True,annot_kws={"size": 16}, fmt='g')
Out[63]:
In [ ]:
2.5 Applying GBDT
2.5.1 Applying XGBOOST on BOW, SET 1¶
In [64]:
# Please write all the code with proper documentation
# merge two sparse matrices: https://stackoverflow.com/a/19710648/4084039
from scipy.sparse import hstack
X_train1=hstack((cat_0n_xtrain, cat_1p_xtrain, sub_cat_0n_xtrain, sub_cat_1p_xtrain, school_state_0n_xtrain,
school_state_1p_xtrain, teacher_prefix_0n_xtrain, teacher_prefix_1p_xtrain,
proj_grade_0n_xtrain, proj_grade_1p_xtrain, price_standardized_xtrain,
teacher_num_prev_projects_standardized_xtrain, quantity_standardized_xtrain,
essay_text_bow_xtrain, proj_title_bow_xtrain)).tocsr()
X_test1=hstack((cat_0n_xtest, cat_1p_xtest, sub_cat_0n_xtest, sub_cat_1p_xtest, school_state_0n_xtest,
school_state_1p_xtest, teacher_prefix_0n_xtest, teacher_prefix_1p_xtest,
proj_grade_0n_xtest, proj_grade_1p_xtest, price_standardized_xtest,
teacher_num_prev_projects_standardized_xtest, quantity_standardized_xtest,
essay_text_bow_xtest, proj_title_bow_xtest)).tocsr()
print(X_train1.shape, y_train.shape)
print(X_test1.shape, y_test.shape)
In [66]:
import dill
dill.dump_session('notebook_6_11.db')
#dill.load_session('notebook_env11.db')
GridSearchCV¶
In [67]:
from sklearn.model_selection import GridSearchCV
import xgboost as xgb
import time
start_time = time.time()
gbdt1 = xgb.XGBClassifier(n_jobs=-1,class_weight='balanced')
parameters = {'n_estimators': [5, 10, 50, 100], 'max_depth':[2, 5, 7, 10]}
clfgbdt1 = GridSearchCV(gbdt1, parameters, cv= 3, scoring='roc_auc',return_train_score=True)
clfgbdt1.fit(X_train1, y_train)
train_auc= clfgbdt1.cv_results_['mean_train_score']
train_auc_std= clfgbdt1.cv_results_['std_train_score']
cv_auc = clfgbdt1.cv_results_['mean_test_score']
cv_auc_std= clfgbdt1.cv_results_['std_test_score']
print("Execution time: " + str((time.time() - start_time)) + ' ms')
In [69]:
train_auc = train_auc.reshape(4,4)
train_auc
Out[69]:
In [70]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(train_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [72]:
cv_auc = cv_auc.reshape(4,4)
cv_auc
Out[72]:
In [73]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(cv_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [74]:
# Testing the performance of the model on test data, plotting ROC Curves
# Select best log(C) value
best_set1_xgb = clfgbdt1.best_params_
print(best_set1_xgb)
In [75]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [76]:
from sklearn.metrics import roc_curve, auc
gbdt1 = xgb.XGBClassifier(max_depth = 5, n_estimators = 100,n_jobs=-1,class_weight='balanced')
gbdt1.fit(X_train1, y_train)
y_train_pred = batch_predict(gbdt1, X_train1)
y_test_pred = batch_predict(gbdt1, X_test1)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(FPR)")
plt.ylabel("True Positive Rate(TPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [77]:
# we are writing our own function for predict, with defined thresould
# we will pick a threshold that will give the least fpr
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(tpr*(1-fpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
return predictions
In [78]:
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
print("Test confusion matrix")
print(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
In [79]:
conf_matr_df_train = pd.DataFrame(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_train, annot=True,annot_kws={"size": 16}, fmt='g')
Out[79]:
In [80]:
conf_matr_df_test = pd.DataFrame(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_test, annot=True,annot_kws={"size": 16}, fmt='g')
Out[80]:
2.5.2 Applying XGBOOST on TFIDF, SET 2¶
In [81]:
# Please write all the code with proper documentation
# merge two sparse matrices: https://stackoverflow.com/a/19710648/4084039
from scipy.sparse import hstack
X_train2=hstack((cat_0n_xtrain, cat_1p_xtrain, sub_cat_0n_xtrain, sub_cat_1p_xtrain, school_state_0n_xtrain,
school_state_1p_xtrain, teacher_prefix_0n_xtrain, teacher_prefix_1p_xtrain,
proj_grade_0n_xtrain, proj_grade_1p_xtrain, price_standardized_xtrain,
teacher_num_prev_projects_standardized_xtrain,
quantity_standardized_xtrain,essay_tfidf_xtrain, proj_title_tfidf_xtrain)).tocsr()
X_test2=hstack((cat_0n_xtest, cat_1p_xtest, sub_cat_0n_xtest, sub_cat_1p_xtest, school_state_0n_xtest,
school_state_1p_xtest, teacher_prefix_0n_xtest, teacher_prefix_1p_xtest,
proj_grade_0n_xtest, proj_grade_1p_xtest, price_standardized_xtest,
teacher_num_prev_projects_standardized_xtest, quantity_standardized_xtest,
essay_tfidf_xtest, proj_title_tfidf_xtest)).tocsr()
print(X_train2.shape)
print(X_test2.shape)
In [82]:
from sklearn.model_selection import GridSearchCV
import xgboost as xgb
import time
start_time = time.time()
gbdt2 = xgb.XGBClassifier(n_jobs=-1,class_weight='balanced')
parameters = {'n_estimators': [5, 10, 50, 100], 'max_depth':[2, 5, 7, 10]}
clfgbdt2 = GridSearchCV(gbdt2, parameters, cv= 3, scoring='roc_auc',return_train_score=True)
clfgbdt2.fit(X_train2, y_train)
train_auc= clfgbdt2.cv_results_['mean_train_score']
train_auc_std= clfgbdt2.cv_results_['std_train_score']
cv_auc = clfgbdt2.cv_results_['mean_test_score']
cv_auc_std= clfgbdt2.cv_results_['std_test_score']
print("Execution time: " + str((time.time() - start_time)) + ' ms')
In [83]:
train_auc = train_auc.reshape(4,4)
cv_auc
Out[83]:
In [84]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(train_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [85]:
cv_auc = cv_auc.reshape(4,4)
cv_auc
Out[85]:
In [86]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(cv_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [101]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [87]:
from sklearn.metrics import roc_curve, auc
gbdt2 = xgb.XGBClassifier(max_depth = 5, n_estimators = 100,n_jobs=-1,class_weight='balanced')
gbdt2.fit(X_train2, y_train)
y_train_pred = batch_predict(gbdt2, X_train2)
y_test_pred = batch_predict(gbdt2, X_test2)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(FPR)")
plt.ylabel("True Positive Rate(TPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [88]:
# we are writing our own function for predict, with defined thresould
# we will pick a threshold that will give the least fpr
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(tpr*(1-fpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
return predictions
In [89]:
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
print("Test confusion matrix")
print(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
In [90]:
conf_matr_df_train = pd.DataFrame(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_train, annot=True,annot_kws={"size": 16}, fmt='g')
Out[90]:
In [91]:
conf_matr_df_test = pd.DataFrame(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_test, annot=True,annot_kws={"size": 16}, fmt='g')
Out[91]:
2.5.3 Applying XGBOOST on AVG W2V, SET 3¶
In [92]:
essay_avg_w2v_vectors_xtrain= np.array(essay_avg_w2v_vectors_xtrain)
proj_title_avg_w2v_vectors_xtrain= np.array(proj_title_avg_w2v_vectors_xtrain)
In [93]:
essay_avg_w2v_vectors_xtest= np.array(essay_avg_w2v_vectors_xtest)
proj_title_avg_w2v_vectors_xtest= np.array(proj_title_avg_w2v_vectors_xtest)
In [94]:
# Please write all the code with proper documentation
# merge two sparse matrices: https://stackoverflow.com/a/19710648/4084039
from scipy.sparse import hstack
X_train3=np.hstack((cat_0n_xtrain, cat_1p_xtrain, sub_cat_0n_xtrain, sub_cat_1p_xtrain, school_state_0n_xtrain,
school_state_1p_xtrain, teacher_prefix_0n_xtrain, teacher_prefix_1p_xtrain,
proj_grade_0n_xtrain, proj_grade_1p_xtrain, price_standardized_xtrain,
teacher_num_prev_projects_standardized_xtrain,
essay_avg_w2v_vectors_xtrain, proj_title_avg_w2v_vectors_xtrain))
X_test3=np.hstack((cat_0n_xtest, cat_1p_xtest, sub_cat_0n_xtest, sub_cat_1p_xtest, school_state_0n_xtest,
school_state_1p_xtest, teacher_prefix_0n_xtest, teacher_prefix_1p_xtest,
proj_grade_0n_xtest, proj_grade_1p_xtest, price_standardized_xtest,
teacher_num_prev_projects_standardized_xtest,
essay_avg_w2v_vectors_xtest, proj_title_avg_w2v_vectors_xtest))
print(X_train3.shape, y_train.shape)
print(X_test3.shape, y_test.shape)
GridsearchCV¶
In [95]:
from sklearn.model_selection import GridSearchCV
import xgboost as xgb
import time
start_time = time.time()
gbdt3 = xgb.XGBClassifier(n_jobs=-1,class_weight='balanced')
parameters = {'n_estimators': [5, 10, 50, 100], 'max_depth':[2, 5, 7, 10]}
clfgbdt3 = GridSearchCV(gbdt3, parameters, cv= 3, scoring='roc_auc',return_train_score=True)
clfgbdt3.fit(X_train3, y_train)
train_auc= clfgbdt3.cv_results_['mean_train_score']
train_auc_std= clfgbdt3.cv_results_['std_train_score']
cv_auc = clfgbdt3.cv_results_['mean_test_score']
cv_auc_std= clfgbdt3.cv_results_['std_test_score']
print("Execution time: " + str((time.time() - start_time)) + ' ms')
In [96]:
train_auc = train_auc.reshape(4,4)
train_auc
Out[96]:
In [97]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(train_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [98]:
cv_auc = cv_auc.reshape(4,4)
cv_auc
Out[98]:
In [99]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(cv_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [102]:
# Testing the performance of the model on test data, plotting ROC Curves
# Select best log(C) value
best_set3_xgb = clfgbdt3.best_params_
print(best_set3_xgb)
In [100]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [103]:
from sklearn.metrics import roc_curve, auc
gbdt3 = xgb.XGBClassifier(max_depth = 5, n_estimators = 50,n_jobs=-1,class_weight='balanced')
gbdt3.fit(X_train3, y_train)
y_train_pred = batch_predict(gbdt3, X_train3)
y_test_pred = batch_predict(gbdt3, X_test3)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(FPR)")
plt.ylabel("True Positive Rate(TPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [106]:
# we are writing our own function for predict, with defined thresould
# we will pick a threshold that will give the least fpr
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(tpr*(1-fpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
return predictions
In [107]:
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
print("Test confusion matrix")
print(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
In [108]:
conf_matr_df_train = pd.DataFrame(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_train, annot=True,annot_kws={"size": 16}, fmt='g')
Out[108]:
In [109]:
conf_matr_df_test = pd.DataFrame(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_test, annot=True,annot_kws={"size": 16}, fmt='g')
Out[109]:
2.5.4 Applying XGBOOST on TFIDF W2V, SET 4¶
In [110]:
essay_tfidf_w2v_vectors_xtrain=np.array(essay_tfidf_w2v_vectors_xtrain)
proj_title_tfidf_w2v_vectors_xtrain=np.array(proj_title_tfidf_w2v_vectors_xtrain)
In [111]:
essay_tfidf_w2v_vectors_xtest=np.array(essay_tfidf_w2v_vectors_xtest)
proj_title_tfidf_w2v_vectors_xtest=np.array(proj_title_tfidf_w2v_vectors_xtest)
In [112]:
# Please write all the code with proper documentation
# merge two sparse matrices: https://stackoverflow.com/a/19710648/4084039
from scipy.sparse import hstack
X_train4=np.hstack((cat_0n_xtrain, cat_1p_xtrain, sub_cat_0n_xtrain, sub_cat_1p_xtrain, school_state_0n_xtrain,
school_state_1p_xtrain, teacher_prefix_0n_xtrain, teacher_prefix_1p_xtrain,
proj_grade_0n_xtrain, proj_grade_1p_xtrain, price_standardized_xtrain,
teacher_num_prev_projects_standardized_xtrain, quantity_standardized_xtrain,
essay_tfidf_w2v_vectors_xtrain, proj_title_tfidf_w2v_vectors_xtrain))
X_test4=np.hstack((cat_0n_xtest, cat_1p_xtest, sub_cat_0n_xtest, sub_cat_1p_xtest, school_state_0n_xtest,
school_state_1p_xtest, teacher_prefix_0n_xtest, teacher_prefix_1p_xtest,
proj_grade_0n_xtest, proj_grade_1p_xtest, price_standardized_xtest,
teacher_num_prev_projects_standardized_xtest, quantity_standardized_xtest,
essay_tfidf_w2v_vectors_xtest, proj_title_tfidf_w2v_vectors_xtest))
print(X_train4.shape, y_train.shape)
print(X_test4.shape, y_test.shape)
GridsearchCV¶
In [113]:
from sklearn.model_selection import GridSearchCV
import xgboost as xgb
import time
start_time = time.time()
gbdt4 = xgb.XGBClassifier(n_jobs=-1,class_weight='balanced')
parameters = {'n_estimators': [5, 10, 50, 100], 'max_depth':[2, 5, 7, 10]}
clfgbdt4 = GridSearchCV(gbdt4, parameters, cv= 3, scoring='roc_auc',return_train_score=True)
clfgbdt4.fit(X_train4, y_train)
train_auc= clfgbdt4.cv_results_['mean_train_score']
train_auc_std= clfgbdt4.cv_results_['std_train_score']
cv_auc = clfgbdt4.cv_results_['mean_test_score']
cv_auc_std= clfgbdt4.cv_results_['std_test_score']
print("Execution time: " + str((time.time() - start_time)) + ' ms')
In [114]:
import dill
dill.dump_session('notebook_71_11.db')
#dill.load_session('notebook_71_11.db')
In [116]:
train_auc = train_auc.reshape(4,4)
train_auc
Out[116]:
In [117]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(train_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [118]:
cv_auc = cv_auc.reshape(4,4)
cv_auc
Out[118]:
In [119]:
import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)
import seaborn as sns
sns.heatmap(cv_auc,annot=True)
plt.xticks(np.arange(4), [2, 5, 7, 10])
plt.yticks(np.arange(4), [5, 10, 50, 100])
plt.xlabel('max_depth')
plt.ylabel('n_estimators')
plt.show()
In [120]:
# Testing the performance of the model on test data, plotting ROC Curves
# Select best log(C) value
best_set4_xgb = clfgbdt4.best_params_
print(best_set4_xgb)
In [121]:
def batch_predict(clf, data):
# roc_auc_score(y_true, y_score) the 2nd parameter should be probability estimates of the positive class
# not the predicted outputs
y_data_pred = []
tr_loop = data.shape[0] - data.shape[0]%1000
# consider you X_tr shape is 49041, then your cr_loop will be 49041 - 49041%1000 = 49000
# in this for loop we will iterate unti the last 1000 multiplier
for i in range(0, tr_loop, 1000):
y_data_pred.extend(clf.predict_proba(data[i:i+1000])[:,1])
# we will be predicting for the last data points
y_data_pred.extend(clf.predict_proba(data[tr_loop:])[:,1])
return y_data_pred
In [122]:
from sklearn.metrics import roc_curve, auc
gbdt4 = xgb.XGBClassifier(max_depth = 5, n_estimators = 50,n_jobs=-1,class_weight='balanced')
gbdt4.fit(X_train4, y_train)
y_train_pred = batch_predict(gbdt4, X_train4)
y_test_pred = batch_predict(gbdt4, X_test4)
train_fpr, train_tpr, tr_thresholds = roc_curve(y_train, y_train_pred)
test_fpr, test_tpr, te_thresholds = roc_curve(y_test, y_test_pred)
plt.plot(train_fpr, train_tpr, label="Train AUC ="+str(auc(train_fpr, train_tpr)))
plt.plot(test_fpr, test_tpr, label="Test AUC ="+str(auc(test_fpr, test_tpr)))
plt.legend()
plt.xlabel("False Positive Rate(FPR)")
plt.ylabel("True Positive Rate(TPR)")
plt.title("AUC")
plt.grid()
plt.show()
In [123]:
# we are writing our own function for predict, with defined thresould
# we will pick a threshold that will give the least fpr
def predict(proba, threshould, fpr, tpr):
t = threshould[np.argmax(tpr*(1-fpr))]
# (tpr*(1-fpr)) will be maximum if your fpr is very low and tpr is very high
print("the maximum value of tpr*(1-fpr)", max(tpr*(1-fpr)), "for threshold", np.round(t,3))
predictions = []
for i in proba:
if i>=t:
predictions.append(1)
else:
predictions.append(0)
return predictions
In [124]:
from sklearn.metrics import confusion_matrix
print("Train confusion matrix")
print(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
print("Test confusion matrix")
print(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
In [125]:
conf_matr_df_train = pd.DataFrame(confusion_matrix(y_train[:], predict(y_train_pred, tr_thresholds, train_fpr, train_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_train, annot=True,annot_kws={"size": 16}, fmt='g')
Out[125]:
In [126]:
conf_matr_df_test = pd.DataFrame(confusion_matrix(y_test[:], predict(y_test_pred, tr_thresholds, test_fpr, test_tpr)))
sns.set(font_scale=1.4)#for label size
sns.heatmap(conf_matr_df_test, annot=True,annot_kws={"size": 16}, fmt='g')
Out[126]:
3. Conclusion
In [ ]:
# Please compare all your models using Prettytable library
In [127]:
from prettytable import PrettyTable
x = PrettyTable()
x.field_names = ["Vectorizer", "Model", "Hyperparameters[max_depth,n_estimators]" , "Test AUC"]
x.add_row(["BOW", "RF","[10,500]", 0.70001])
x.add_row(["TFIDF", "RF", "[10,100]", 0.6903])
x.add_row(["AVG W2V", "RF", "[10,10]", 0.61638])
x.add_row(["TFIDF W2V", "RF", "[10,100]", 0.69172])
x.add_row(["BOW", "GBDT","[5,100]", 0.71415])
x.add_row(["TFIDF", "GBDT", "[5,100]", 0.71747])
x.add_row(["AVG W2V", "GBDT", "[5,50]", 0.6825])
x.add_row(["TFIDF W2V", "GBDT", "[5,50]", 0.70865])
print(x)
In [ ]: